In silico approaches for pathogenicity prediction of missense variants of uncertain significance (VUS) in ADPKD

Abstract

Autosomal dominant polycystic kidney disease (ADPKD) is the most common dominantly inherited kidney disease, being caused by mutations in PKD1 (85%) and PKD2 (15%). Approximately 25% of all PKD1 mutations are pathogenic missense,      but additional 12.8% of those are reported as indeterminate or missense variants of uncertain significance (VUS), of which the deleterious nature are unclear in clinical practice. Functional studies to assess the impact of missense variants, particularly  for PKD1, are difficult due to the large size and ambiguous functions of the proteins, polycystins. A variety of in silico tools were developed to evaluate interspecies variations and biochemical impacts of amino acid substitutions of missense VUS. To evaluate the tool suitability for PKD1 and PKD2, we applied additional 12 state-of-the-art web-based tools with their claimed superior performances based on different algorithms for the prediction in parallel with the three benchmark programs (PolyPhen2, SIFT and MutationTaster). A total of 15 tools were assessed in a gene-specific manner with PKD1 and PKD2 variants of known pathogenicity from the PKDB mutation database. We found that each of the genes had suitable predictions from different sets of tools. Combined uses of 3, 5, 8 and 12 tools with high performance in descending order (in which the benchmarks were excluded) gave consensus predictions for the selected nine VUS in PKD1, except one VUS which might have a mild pathogenicity that only the use of 12 tools predicted differently. In this situation, using more numbers of tools might prevent the misinterpretation of milder VUS. Classification of the pathogenicity of the VUS in PKD1 and PKD2 became an essential part of molecular diagnosis of ADPKD and is useful for a clinical decision and would expand knowledge of mutation spectrum in ADPKD.